| This paper relies on the national "Eleventh Five-Year"863-project named"Intelligent Cleaning Robot for Large-scale Power Plant Condenser"(Item Number:2008AA04Z214). In order to meet the need of the600MW generator-sets’ condensercleaning operation in Datang Leiyang Thermal Power-Plant, the visual positioningtechnology of mobile cleaning robot is researched. After introducing the mechanicalstructure and electrical motion control system of the robot body briefly, this paperfocuses on the visual object-localization algorithm of the condenser cleaning mobilerobot in complex environment, in order to improve the efficiency of the condensercleaning, reduce labor intensity, and bring great socio-economic benefits.Firstly, the mechanism and features of many types of mobile robots based onvisual localization technology are integrated and analyzed, combined with theenvironment characteristics and cleaning operations requirement of the largecondenser equipment, the necessity of the design and development of thethird-generation prototype of the condenser cleaning mobile robot is clarified.Secondly, a brief analysis of the automatic condenser cleaning equipment isdescribed, including the feature of the robot’s track-wheel-driven mobile joints, andthe characteristics of the high pressure water cleaning gun.Thirdly, because the working environment of the condenser is unknown, dirtynozzle regions of the tube plate is uncertain and the contours’ feature of dirty nozzleis not complete, it is necessary to design a visual localization program for thecondenser cleaning mobile robot to achieve vision–guiding of spray gun at the targetlocation. The visual subsystem of the prototype has been improved with multi-anglefixed lights and vision sensor-based observation equipment, supplemented byultrasonic ranging sensor system and network communication system, by cameracalibration and environmental characteristic matching methods, to realize the remotecontrolling of robot’s cleaning-side movement to locate dirty area of the heatexchanger tube.Again, in order to initially determine the dirty nozzle region, this paper presentsan algorithm, based on cumulative statistical laws of the row-column pixel values ofimages to recognize and locate the multi-nozzle objects. To meet different locatingaccuracy requirement of sampling density, this paper presents an adjustable contour fitting location algorithm, to precisely locate the centroids of nozzle. By calculatingthe transformation matrix from the two-dimensional view coordinate system to thethree-dimensional world coordinate system of target feature point, the coordinates ofthe nozzle centroids in the world coordinate system can be determined, and accuratepositional parameters for the trajectory planning and motion error compensation ofthe spray gun can be obtained.Finally, this paper gives a summary of the finished work and achievement, andproposes the future research direction. |
PDF Full Text Request